Vibroacoustic sensorics for medical device guidance

我们建议使用医疗介入设备（如针头、活检系统、导管、抓握器、介入工具）与人体组织相互作用时产生的振动声信号来增加有关临床事件的有价值信息，提供定向指导，从而减少相关的医疗错误。我们在该领域的先前工作表明，进入新组织的设备尖端（远端）在很大程度上负责近端接收到的音频信号，并且进入不同的组织会改变音频特征。到目前为止，我们已经设法在实验室设置中获得数据点，使用连接到近端的初始传感器设计，从而消除了开发专用临床干预工具和电缆的需要，并减少了不育问题。这项技术可能不会对现有的和经过验证的临床工具进行重大改变。本提案的研究工作将尝试将该实验室技术引入特定的基于针的临床应用（进入股血管-静脉和动脉），从而提高手术质量。目的构建的近端MEMS音频传感器将数据传输到采集系统，该系统对信号进行预处理，提取相关特征（高级信号处理），并尝试表征（机器学习）干预期间的重要事件，如穿过主要组织层或进入和离开血管，并给出指向目标目的地的定向信息，并确认已达到医疗目标。我们将与临床工作人员密切互动，记录实际手术过程、手术过程中的互动以及其他相关事件，作为系统建模的基础。理解和模拟信号及其行为的一个重要发展步骤是建立一个专用的合成体来模拟和匹配从进入动物组织样本中获得的先前结果的发现。这与医疗程序信息一起用于建立一个分类模型，该模型将允许我们基于诊断成像分析（超声）模拟人类干预。该装置随后将在临床介入放射设置中使用正常导管过程中获得的真实患者数据进行测试、校正和优化。下一步是通过表明外科医生性能的提高（瞄准精度、手术速度、置信度）来验证所提出的指导方法。这项研究在其他相邻和相关的研究领域具有巨大的潜力，可能会提高制导精度，基于音频信号的不同软组织表征，以及为未来（半）自主机器人手术/导管介入提供额外的引导信息和模拟触觉反馈/事件信息。

We propose to use vibroacoustic signals that are generated during the interaction of medical interventional devices (e.g. needles, biopsy systems, catheters, graspers, interventional tools) with human tissue to add valuable information about clinical events, provide directional guidance and subsequently reduce related medical errors. Our prior work in that field has shown that the device tip (distal end) entering new tissue is to a large extent responsible for an audio signal received on the proximal end and that entering different tissues changes the audio characteristics. So far, we have managed to obtain data points in a laboratory setup using an initial sensor design attached to the proximal end and by that eliminating the need to develop dedicated clinical intervention tools, cables and reducing sterility issues. This technology likely will work without major alteration to existing and proven clinical tools. The research work of this proposal will attempt to introduce this laboratory technology to a specific needle-based clinical application (access to the femoral vessels - venous as well as arterial) that could lead to improve the quality of the procedure. A purpose build proximal MEMS audio sensor will transfer the data to an acquisition system that pre-processes the signal, extracts relevant features (advanced signal processing), and attempts to characterize (machine learning) important events during the intervention, like crossing main tissue layers or entering and leaving the vessel, and to give directional information towards the target destination and a confirmation on having reached the medical goal. We will closely interact with the clinical staff and record the actual procedure, interactions during the procedure, and other relevant events as a base for the system modeling. An important development step to understand and simulate signals and their behaviors is to build a dedicated synthetic phantom to model and match the findings of prior acquired results from entering animal tissue samples. This - together with the medical procedure information - is used to build a classification model that will allow us to simulate human interventions based on diagnostic imaging analysis (ultrasound). The setup will subsequently be tested, corrected, and optimized with real patient data obtained during a normal catheter procedure in a clinical interventional radiology setup. The next step is to validate the proposed guidance methodology in phantoms by indicating the increase in surgeonsperformance (targeting accuracy, speed of the procedure, confidence level. This research has enormous potential in other adjacent and connected research areas possibly allowing improved guidance accuracy, audio signal-based characterization of different soft tissues, as well as to provide additional guidance information and simulated haptic feedback/event information for a future (semi-)autonomous robotic surgery/catheter intervention.